[
https://issues.apache.org/jira/browse/SPARK-761?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14491094#comment-14491094
]
Andrew Ash commented on SPARK-761:
----------------------------------
An implementation I can imagine would be to do an exchange of the Spark binary
versions (e.g. 1.2.2, 1.3.1, 1.4.0, etc) on startup and log errors.
The catch though is that Spark maintains forward compatibility between
versions, so when used with the spark-submit script, a 1.2.2-compiled job jar
will run on a 1.3.1 cluster. So we don't want to do a compatibility check on
the Spark binary versions directly, but rather on the API that they speak,
which is not versioned independently of the Spark binaries.
> Print a nicer error message when incompatible Spark binaries try to talk
> ------------------------------------------------------------------------
>
> Key: SPARK-761
> URL: https://issues.apache.org/jira/browse/SPARK-761
> Project: Spark
> Issue Type: Improvement
> Components: Spark Core
> Reporter: Matei Zaharia
> Priority: Minor
> Labels: starter
>
> As a starter task, it would be good to audit the current behavior for
> different client <-> server pairs with respect to how exceptions occur.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
